The Research Of Application And Methods Of The Image Visual Salience Detection

Due to the arrival of information era and rapid development of internet technologies, we can receive massive data via internet and other media every day. In recent years, with the maturity of storage technology, information transfers faster and faster, hardware’s storage capacity become bigger and bigger, which cause we exposed to an increasing number of video image information. For image data, the human visual system can spontaneously focus on the most important area in the image, the visual function of human is called visual saliency, and area of visual attention is saliency region. This paper mainly studies the visual saliency detection algorithm, which can quickly find saliency region form an image and extract the saliency map.It has formed relevant theoretical achievements and lots of saliency detection algorithms have been proposed through the efforts of many researchers. In this paper, we propose two saliency detection algorithms and apply it to image enhancement, the main work has the following components:(1) Summarize the current significant theoretical knowledge. Discuss the saliency detection problem respectively form aspects of visual saliency, saliency features, algorithm classification, and analyze several classical saliency detection algorithms.(2) Propose an algorithm based on image synthesis contrast. According to the existing two types of saliency detection methods: local contrast algorithm and global algorithm, we propose an image algorithm synthesize local features and global features. This algorithm combines the classic local contrast algorithm(AC algorithm) and global contrast algorithm(HC algorithm), and on this basis, make a space weight and regional optimization to the output saliency image, which add spatial information and enhance the contrast of the output image.(3) Propose an algorithm based on image depth information. As for complex image containing depth information, we propose an algorithm combined depth information. This algorithm distinguish the near and far regions of the original image, using super-pixel segmentation method(SLIC) and Markov random field model to extract depth map, on the basis of classical frequency tuned algorithm(FT algorithm), combination the depth map readjust the pixel saliency value to obtain the output saliency image.(4) Introduce the application of saliency detection algorithm in the image enhancement. In the view of the backlit face image, the saliency detection method was adopted for image enhancement processing, this method only enhance the saliency face region, while do not processing the background region with normal illumination intensity.